Professional Documents
Culture Documents
James B. Rawlings
Department of Chemical and Biological Engineering
University of WisconsinMadison
Insitut f
ur Systemtheorie und Regelungstechnik
Universitat Stuttgart
Stuttgart, Germany
June 20, 2011
1 / 46
Outline
Introduction
2 / 46
Predictive control
Reconcile the past
11
00
00
11
0
1
0
1
0
1
11
00
00
11
00
11
MH Estimate
Measurement
0
1
0
1
1
0
1
0
1
0
0
1
11
0
1
0
0
1
0
1
0
1 1
0
01
1
0
0
1
0
0
1
0
01
0 0
01
1
0
1
1 1
0
1
1
0
1
0
sensors
y
1
0
1
0
1
0
0
1
0
1
0
1
1
0
1
0
1
0
0
1
1
0
0
1
1
0
1
0
1
0
0
1
1
0
1
0
1
0
1
0
01
1
0
01
0
11
0
11
0
0
0
1
1
0
0
1
1
0
0
1
10
0
1
0
1
0
1
0
11
0
0
1
0
1
1
0
1
0
1
0
0
1
1
0
0
1
1
0
1
0
1
0
0
1
1
0
1
0
1
0
1
0
1
0
0
1
Forecast
MPC control
1
0
0
0
0
0
0
0
0
0
0
0
0
01
1
01
1
01
1
01
1
01
1
01
1
01
1
01
1
01
1
01
1
01
1
0
1
1
0
0
1
1
0
actuators
u
time
min
u(t)
x = f (x, u)
x(0) = x0
(given)
y = g (x, u)
Stuttgart June 2011
3 / 46
State estimation
Reconcile the past
11
00
00
11
0
1
0
1
0
1
sensors
y
0
1
0
1
1
0
1
0
1
0
0
1
11
0
1
0
0
1
0
1
0
1 1
0
01
1
0
0
1
0
0
1
0
01
0 0
01
1
0
1
1 1
0
1
1
0
1
0
1
0
0
1
MH Estimate
Measurement
1
0
1
0
1
0
0
1
0
1
0
1
1
0
1
0
1
0
0
1
1
0
0
1
1
0
1
0
1
0
0
1
1
0
1
0
1
0
1
0
0
1
0
01
1
0
1
1
0
11
0
0
0
1
1
0
0
1
10
0
1
0
1
0
1
1
0
0
1
1
0
1
0
0
11
0
0
1
0
1
1
0
1
0
1
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
0
0
1
Forecast
MPC control
1
0
0
0
0
0
0
0
0
0
0
0
0
01
1
01
1
01
1
01
1
01
1
01
1
01
1
01
1
01
1
01
1
01
1
0
1
1
0
0
1
1
0
actuators
u
t
min
x0 ,w (t) T
time
x = f (x, u) + w
(process noise)
y = g (x, u) + v
(measurement noise)
4 / 46
System model1
We consider systems of the form
x + = f (x, u)
where the state x lies in Rn and the control (input) u lies in Rm ;
In this formulation x and u denote, respectively, the current state and
control, and x + the successor state.
We assume in the sequel that the function f : Rn Rm Rn is
continuous.
Let
(k; x, u)
denote the solution of x + = f (x, u) at time k if the initial state is
x(0) = x and the control sequence is u = {u(0), u(1), u(2), . . .};
The solution exists and is unique.
1
Most of this preliminary material is taken from Rawlings and Mayne (2009,
Appendix B). Downloadable from www.che.wisc.edu/~jbraw/mpc.
Stuttgart June 2011
5 / 46
6 / 46
7 / 46
8 / 46
9 / 46
Definition 4
A function : R0 R0 belongs to class K if it is continuous, zero
at zero, and strictly increasing;
: R0 R0 belongs to class K if it is a class K and unbounded
((s) as s ).
A function : R0 I0 R0 belongs to class KL if it is
continuous and if, for each t 0, (, t) is a class K function and for
each s 0, (s, ) is nonincreasing and satisfies limt (s, t) = 0.
A function : R R0 belongs to class PD (is positive definite) if it
is continuous and positive everywhere except at the origin.
10 / 46
11 / 46
Stability Definitions
locally stable in X if, for each > 0, there exists a = () > 0 such
that x X (A B), implies |(i; x)|A < for all i I0 .a
12 / 46
13 / 46
i I0
(1)
See Teel and Zaccarian (2006) and the Notes on Recent MPC
Literature link on: www.che.wisc.edu/~jbraw/mpc for further
discussion of the differences in the two definitions.
If f () is continuous, the two definitions are equivalent.
Stuttgart June 2011
14 / 46
Lyapunov function
Definition 7 (Lyapunov function)
A function V : Rn R0 is said to be a Lyapunov function for the system
x + = f (x) and set A if there exist functions i K , i = 1, 2 and
3 PD such that for any x Rn ,
V (x) 1 (|x|A )
(2)
V (x) 2 (|x|A )
(3)
(4)
15 / 46
Remark 8
It is shown in Jiang and Wang (2002), Lemma 2.8, that, under the
assumption that f () is continuous, we can always assume that 3 () in (4)
is a K function.
16 / 46
17 / 46
18 / 46
19 / 46
20 / 46
Assume that the origin is globally exponentially stable (GES) for the
system
x + = f (x)
in which f is Lipschitz continuous. Show that there exists a Lipschitz
continuous Lyapunov function V () for the system satisfying for all
x Rn
a1 |x| V (x) a2 |x|
V (f (x)) V (x) a3 |x|
in which a1 , a2 , a3 , > 0.
Hint: Consider summing the solution |(i; x)| on i as a candidate
Lyapunov function V (x).
Establish also that in the Lyapunov function defined above, any > 0
is valid, and the constant a3 can be chosen as large as one wishes.
Stuttgart June 2011
21 / 46
Robust Stability
We turn now to stability conditions for systems subject to bounded
disturbances (not vanishingly small) and described by
x + = f (x, w )
(5)
(6)
22 / 46
We require, in the sequel, that the set A is positive invariant for (5) (or
for x + F (x)):
23 / 46
24 / 46
Asymptotic stability
25 / 46
Lyapunov function
To cope with disturbances we require a modified definition of a Lyapunov
function.
(7)
V (x) 2 (|x|A )
(8)
(9)
zF (x)
26 / 46
GAS (disturbances)
27 / 46
Input-to-State Stability
In input-to-state stability (Sontag and Wang, 1995; Jiang and Wang,
2001) we seek a bound on the state in terms of a uniform bound on the
disturbance sequence w := {w (0), w (1), . . .}. Let kk denote the usual `
norm for sequences, i.e., kwk := supk0 |w (k)|.
28 / 46
(10)
u(k) U
for all k I0
29 / 46
Feasible sets
The set of feasible initial states and associated control sequences
ZN = {(x, u) | u(k) U, (k; x, u) X for all k I0:N1 ,
and (N; x, u) Xf }
and Xf X is the feasible terminal set.
The set of feasible initial states is
XN = {x Rn | u UN such that (x, u) ZN }
(11)
30 / 46
N1
X
k=0
min VN (x, u)
uUN
31 / 46
difference equation
x + f (x, N (x))
difference inclusion
32 / 46
w = (d, e)
33 / 46
34 / 46
x XN , u U
x Xf
35 / 46
We show that the optimal cost VN0 () is a Lyapunov function for the
closed-loop system. We require three properties.
Lower bound.
VN0 (x) 1 (|x|)
for all
x XN
for all
x XN , u UN
36 / 46
for all
x XN
37 / 46
x Xf
so we have that
) VN0 (x) `(x, u(0; x))
VN (x + , u
The optimal cost is certainly no worse, giving
VN0 (x + ) VN0 (x) `(x, u(0; x))
VN0 (x + ) VN0 (x) 1 (|x|)
for all
x XN
38 / 46
Upper bound
Finally we require the upper bound.
VN0 (x) 2 (|x|)
for all
x XN
for all
x Xf
Vf (x)
x Xf
39 / 46
x X2
x X2
x X1
x X1
Therefore
V20 (x) Vf (x) x Xf
Continuing this recursion gives for all N 1
VN0 (x) Vf (x)
x Xf
40 / 46
for all
x XN
Be aware that the MPC literature has been confused about the
requirements for this last result.
2
41 / 46
42 / 46
Recommended exercises
43 / 46
Further Reading I
H. Chen and F. Allg
ower. A quasi-infinite horizon nonlinear model
predictive control scheme with guaranteed stability. Automatica, 34(10):
12051217, 1998.
Z.-P. Jiang and Y. Wang. Input-to-state stability for discrete-time
nonlinear systems. Automatica, 37:857869, 2001.
Z.-P. Jiang and Y. Wang. A converse Lyapunov theorem for discrete-time
systems with disturbances. Sys. Cont. Let., 45:4958, 2002.
H. K. Khalil. Nonlinear Systems. Prentice-Hall, Upper Saddle River, NJ,
third edition, 2002.
J. B. Rawlings and D. Q. Mayne. Model Predictive Control: Theory and
Design. Nob Hill Publishing, Madison, WI, 2009. 576 pages, ISBN
978-0-9759377-0-9.
E. D. Sontag and Y. Wang. On the characterization of the input to state
stability property. Sys. Cont. Let., 24:351359, 1995.
Stuttgart June 2011
44 / 46
Further Reading II
45 / 46
Acknowledgments
46 / 46